Urban Ecosystems (2020) 23:905–917 https://doi.org/10.1007/s11252-020-00988-2 Crown width models for woody plant species growing in urban areas of the U.S. James A. Westfall1 & David J. Nowak2 & JasonG.Henning3 & Tonya W. Lister1 & Christopher B. Edgar4 & Mark A. Majewsky5 & Nancy F. Sonti6 Published online: 19 March 2020 # This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2020 Abstract Crown widths of woody plant species growing in urban areas are of considerable importance as an overall indicator of health and also serve as an important factor for assessing leaf area and associated ecosystem services, such as carbon sequestration, air pollution removal, air temperature cooling, and rainfall interception. Unfortunately, assessing crown widths in urban environ- ments is often challenging and time consuming. To help reduce data collection costs and provide consistency over time, models to predict crown widths for urban-grown species were developed using data from 49 cities across the U.S. and Southern Canada. The effort consisted of fitting mixed models for 29 species groups that encompassed 964 species. Cities were considered a random effect and were statistically significant for 22 of the 29 groups. The need for urban-specific crown width models was demonstrated via examination of prediction biases found when applying crown width models based on forest grown trees, where under-prediction up to about 20% was found for the same species growing in urban areas. Application of the models was evaluated by using crown width predictions instead of observed values for calculations of crown leaf area. Mean percent differences in leaf area were about ±10% across most species groups. Further improvements to national-scale urban crown width models should be pursued as additional data become available via i-Tree, Urban FIA, and possibly other sources where data collection protocols are compatible. Keywords Ecosystem services . Forest inventory . Mixed models . Leaf area . Spatial trend Introduction characteristics of forest trees, for example, have been actively studied due to their high correlation with tree growth (Chen As forest inventories of urban areas become more common- et al. 2017; Leites et al. 2009), likelihood of mortality place, much research is needed to understand phenomena that (Bussotti and Pollastrini 2017; Morin et al. 2015), probability have only previously been studied in forested settings. Crown and behavior of crown fire (Hevia et al. 2018;Mitsopoulos and Dimitrakopoulos 2007), and functional benefits such as air pollution removal (Nowak et al. 2014; Smith 1990). In Electronic supplementary material The online version of this article (https://doi.org/10.1007/s11252-020-00988-2) contains supplementary urban settings, tree crown measurements of woody plant spe- material, which is available to authorized users. cies (hereafter referred to as ‘trees’ for simplicity) are primar- ily used to assess crown size and leaf area, and consequent * James A. Westfall ecosystem services such as carbon sequestration, air pollution [email protected] removal, air temperature cooling, and rainfall interception (Willis and Petrokofsky 2017; Kardan et al. 2015). Urban 1 U.S. Forest Service, Northern Research Station, York, PA, USA corollaries to typical uses of crown information from forested 2 trees include prediction of individual-tree growth and mortal- U.S. Forest Service, Northern Research Station, Syracuse, NY, USA 3 ity (Nowak et al. 2008;Vogtetal.2015). Thus, assessments of Davey Institute, Philadelphia, PA, USA crown dimensions and condition play a pivotal role in under- 4 University of Minnesota, St. Paul, MN, USA standing tree functional processes and their interaction with 5 U.S. Forest Service, Northern Research Station, Tofte, MN, USA urban environmental conditions. 6 U.S. Forest Service, Northern Research Station, Baltimore, MD, As with forest-grown trees, urban tree crown attributes are USA heavily influenced by their local environment. Crown width is 906 Urban Ecosyst (2020) 23:905–917 one important and common tree measurement found in urban 2008; i-Tree 2019). Typically, data are collected within ran- forest inventory protocols. Key factors affecting crown width domly located circular plots having a11.34 m radius. Site include amount of growing space (Pretzsch et al. 2015)and measurements include vegetation and other ground cover water/nutrient availability (Gaudio et al. 2017). In forested types, as well as land use characteristics. On each plot, woody micro-environments, these factors are primarily constrained plants with a minimum stem diameter of 2.54 cm at a height of by competition from neighboring trees (Sharma et al. 2016; 1.37 m are recorded as trees. For each tree, species, diameter, Bragg 2001). Growing space for urban trees may also be af- height, crown width, height to crown, crown light exposure, fected by nearby trees, as well as buildings and other above- and amount of missing crown are recorded (i-Tree 2019). ground formations that place limitations on light availability Additional data (20%) arise from urban inventories initiat- (Tan and Ismail 2015). These factors, along with surface and ed by the Forest Inventory and Analysis (FIA) program of the below-ground conditions, likely also play a role in water and U.S. Forest Service during 2014–2017. The plot design con- nutrient availability that affect tree growth (Clark and sists of a primary circular plot of 14.63 m radius, within which Kjelgren 1990; Berrang et al. 1985). Tree crowns in urban are four microplots having 2.07 m radius located in each car- areas may also be subjected to manipulations such as pruning dinal direction at 3.66 m from plot center (Fig. 1). A myriad of or other types of damage that produce crown sizes not strictly site-level data are collected, such as land use, canopy cover, controlled by natural ecophysiological processes (Fini et al. vegetation cover, and surface cover (U.S. Forest Service 2015; Christie and Hochuli 2005). 2017). Sample trees having diameter (breast-height or root- Despite the numerous factors that affect crown width, sta- collar depending on species) ≥ 12.70 cm were tallied on the tistical models can be developed that predict crown width primary plot; whereas sample trees with diameter ≥ 2.54 cm using tree attribute information. As may be expected, efforts and < 12.70 cm are recorded on microplots. Tree-level data have traditionally focused on forest-grown trees where vari- collection includes measurements of species, diameter, height, ables such as tree diameter, total height, height-diameter ratio, crown width, crown ratio, crown light exposure, and crown height to crown, and crown ratio have been shown to be im- dieback (U.S. Forest Service 2017). portant predictors (Sharma et al. 2016;Fuetal.2013; The i-Tree and Urban FIA data were combined to provide Bechtold 2003). Some models specific to crown width predic- information covering 49 cities across the conterminous U.S. tion for urban trees have also been presented, with tree diam- and southern Canada (Fig. 2). The Canadian cities of Calgary eter being the primary predictor variable (Pretzsch et al. 2015; and Toronto were included because of their close proximity to Troxel et al. 2013; Peper et al. 2001). The use of additional the U.S. border. Due to the large number of species present predictor variables has produced mixed success, with some (964), species groups that generally reflect those developed by evidence that outcomes may be species-dependent (Blood the FIA program (U.S Forest Service 2015)wereusedtofa- et al. 2016). In the U.S., McPherson et al. (2016) developed cilitate analysis (Online Resource 1). A summary of the 29 crown width models for street trees using tree diameter as the species groups is given in Table 1. sole predictor from data encompassing 171 species across 17 A limitation of the combined data is that only variables cities. However, much additional urban tree data exists from collected and having identical definition in common to both across the urban landscape to allow for considerable expan- sources can be utilized. As will be discussed in the subsequent sion of both species inclusion and spatial resolution in model- section, the primary variables that were ultimately of interest ing efforts. Specifically, the objectives of this study are: 1) in this study were crown width, diameter, and total height. develop crown width prediction models applicable to most Distribution statistics for these variables by species group woody species in urban environments of the U.S., 2) establish are shown in Table 2. appropriate uncertainty statistics in light of inconstant vari- ance and correlated observations, 3) account for species group Analysis and city location effects in model calibration, and 4) evaluate the use of model predictions in calculations of urban tree leaf There were few variables in common to both data sources. area. Apart from the primary variables crown width, diameter, and height, the only relevant tree-level variable for consideration in model development was crown light exposure (CLE, U.S. Methods Forest Service 2017). CLE describes the amount of sunlight the crown receives as categorized by 6 levels ranging from Data no light to full light (open grown). It was also possible to ascertain if a tree was measured in an area that was consid- The data used in this study arise from two sources. The data ered to function as a forest (FOR, U.S. Forest Service 2017). are primarily (80%) composed of measurements taken in ur- In these cases, it might be expected that trees would have ban inventories following i-Tree protocols (Nowak et al. characteristics similar to those growing in rural forest Urban Ecosyst (2020) 23:905–917 907 Fig. 1 Urban plot design used by the FIA program Fig. 2 Map of cities in the U.S.